Last month I shipped a UX redesign that cut a therapy app's booking flow from 7 steps to 3. First-time completion went from 34% to 96% in usability testing, and to 62% in production. The team had assumed the problem was length. The actual problem was that we were asking the wrong things at the wrong time.
Here's how the 6-week sprint actually went.
The hypothesis we walked in with was wrong
The product team at Take Therapy had a clear theory before I joined: "Onboarding is too long. Users drop off because we ask too much." It's a reasonable theory. The funnel data supported it. Support tickets confirmed it. Stakeholders were aligned.
I almost shipped to that hypothesis.
In week 1 I ran 14 user interviews — recruited via support email and UserTesting, all first-time users within their first 7 days post-signup. Six themes emerged. The dominant one wasn't "this is too long." It was: "why are you asking me about insurance before I've even met a therapist?"
That reframing changed everything.
What people actually needed
The data was specific:
- 12 of 14 users said urgency was their top driver. Setting therapist preferences upfront felt premature, even ridiculous, when they didn't yet know what therapy would feel like for them.
- 41% of all drop-offs happened at the insurance verification step. Of those who dropped, 60% later returned and paid out-of-pocket.
- Time-to-first-session correlated 3× more strongly with week-2 retention than any other variable. The faster a user got to that first conversation, the more likely they were to stay.
We weren't asking too much. We were asking the wrong things at the wrong moment.
The principle that drove the redesign
I named the design principle in week 3 and pinned it to every Figma frame for the rest of the sprint:
Speed over precision. Get them to a session. Optimise the profile later.
That principle gave us a binary test for every form field: does this question need to be answered before the first session is booked, for the booking transaction or clinical safety? If no, defer it.
The result: 9 of the original 12 onboarding fields moved to a post-first-session "Complete your profile" loop, gamified with a progress meter. Insurance moved to optional. Therapist preferences became something users could rate-and-refine after their first call.
What we shipped
Three screens:
- Concern picker — 6 illustrated chips covering 92% of first-session intents. Required for clinical safety routing (this one we couldn't defer).
- Availability + therapist match — Picks a slot, matches to the next-available therapist who fits the concern. No preferences yet.
- Confirmation — One screen showing therapist photo, slot time, what to expect. Sticky "Add to calendar" button.
That's it. Account creation is implicit (email + password before the concern picker). Insurance, preferences, history — all post-first-session.
What I'd do differently
A few things I caught too late:
- I should have recruited usability test participants in week 1, not week 5. I scrambled in the last week to assemble a cohort of 8.
- I should have brought engineering into the research synthesis. Bilal, our senior iOS engineer, mentioned in week 5 that we could pre-compute insurance verification asynchronously. That comment would have been worth a week earlier in week 1.
- I should have explicitly designed the "what to test next" backlog as part of handoff. I left it implicit. Three months later, the team is still using the redesign, but the planned experimentation loop hasn't shipped.
Senior maturity is partly noticing the things you wish you had done differently and writing them down.
What this work taught me
That the hypothesis the team brings you is rarely the actual problem. The actual problem is usually one or two clicks deeper. Real user research, even at small N, surfaces it fast.
That defining success metrics in week 3 instead of week 6 lets you self-validate decisions without escalating to PM. I stopped needing to ask "is this good?" because I could just check the metric I was driving toward.
That progressive disclosure is one of the most under-used principles in UX. We default to forms because forms are easy to build. Cards are work. But cards are also what users actually want.
The Take Therapy case study goes deeper into the research, the journey map, the iteration timeline, and the design system contributions. If you're redesigning an onboarding flow yourself, that's where to dig in.
If you have a flow that needs the same treatment, tell me about your project.